Tumor markers in rheumatoid arthritis -associated interstitial lung disease: a systematic review and meta-analysis Authors: Zeng Qiaoqian, Wu Qing, Li Bowen, Wei Qijiao, Wang Na Published: 2026 ABSTRACT AbstractObjectivesInterstitial lung disease (ILD) is a serious extra-articular complication of rheumatoid arthritis (RA) and a leading cause of mortality. This meta-analysis aims to comprehensively evaluate the diagnostic utility and clinical relevance of tumor-associated biomarkers in patients with Rheumatoid Arthritis-Associated Interstitial Lung Disease (RA-ILD).MethodsA systematic literature search was conducted in PubMed, EMBASE, Web of Science, CBM, Wan-Fang, and CNKI databases from inception to April 1, 2025. We included case-control and cohort studies that compared tumor marker levels between RA-ILD and RA-non-ILD patients, following the PICOS framework. Study quality was assessed using the Newcastle-Ottawa Scale (NOS). Pooled effect sizes were calculated as standardized mean differences (SMD) with 95% confidence intervals (CI) using random-effects models. Diagnostic accuracy was evaluated using summary receiver operating characteristic (ROC) curves.ResultsThirty-six studies comprising 7300 RA patients were included. Compared to RA-non-ILD patients, those with RA-ILD showed significantly elevated levels of CA125 (SMD = 1.10, 95%CI 0.47–1.72), CA199 (SMD = 0.90, 95%CI 0.56–1.23), CA153 (SMD = 1.70, 95%CI 1.16–2.23), CEA (SMD = 0.58, 95%CI 0.24–0.92), KL-6 (SMD = 1.68, 95%CI 1.22–2.14), NSE (SMD = 0.32, 95%CI 0.16–0.48), and HE4 (SMD = 1.87, 95%CI 1.60–2.14). After adjustment for publication bias using trim-and-fill analysis, CYFRA21-1 was also significantly elevated (SMD = 0.48, 95%CI 0.06–0.89). For distinguishing RA-ILD, KL-6 demonstrated the highest diagnostic accuracy (AUC = 0.964), followed by CYFRA21-1 (AUC = 0.906).ConclusionOur findings, which are primarily based on evidence from Chinese cohorts, support the utility of CA125, CA199, CA153, CEA, KL-6, NSE, HE4, and CYFRA21-1 as biomarkers for RA-ILD. The superior diagnostic performance of KL-6 and CYFRA21-1 suggests their potential clinical utility as part of a multi-marker panel discriminating RA-ILD from RA-non-ILD and risk stratification, offering a complementary approach to imaging for identifying this high-risk RA complication.Registration numberPROSPERO CRD42022362487.Supplementary InformationThe online version contains supplementary material available at https://doi.org/10.1186/s41927-026-00664-6.Keywords: Rheumatoid arthritis, Rheumatoid arthritis-associated interstitial lung disease, Tumor markers, Systematic review, Meta-analysis Introduction Rheumatoid arthritis (RA) is a systemic autoimmune disease characterized by chronic joint inflammation, with interstitial lung disease (ILD) representing one of the most severe extra-articular manifestations and a leading cause of mortality [1]. Epidemiological studies indicate that ILD affects RA patients, and its presence significantly reduces 5-year survival rates [2]. While high-resolution computed tomography (HRCT) remains the gold standard for diagnosing RA-ILD [3], its sensitivity for detecting early-stage pulmonary interstitial lesions is limited, and it fails to provide molecular insights into underlying pathophysiological mechanisms. Furthermore, subclinical pulmonary damage may precede detectable imaging abnormalities, underscoring the urgent need for early biomarkers. Recent studies have highlighted the aberrant expression of tumor markers in RA-ILD [4]. Despite promising associations between tumor markers and RA-ILD, the diagnostic utility of individual markers remains suboptimal, and a comprehensive synthesis of the evidence is lacking. This meta-analysis aims to synthesize existing clinical evidence to assess the diagnostic accuracy of tumor markers in differentiating RA-ILD from RA without ILD. Our findings may inform the development of multidimensional early-detection strategies to optimize clinical management of RA-ILD. Materials and methods Literature-search strategy A comprehensive literature search was conducted in several databases, including MEDLINE/PubMed, EMBASE, Web of Science, the Chinese Biomedical Literature Database (CBM), the Wan-Fang Database, and the China National Knowledge Infrastructure (CNKI). The search was performed from database inception until April 1, 2025, with no language restrictions. The search strategy combined terms for rheumatoid arthritis, interstitial lung disease, and specific tumor-associated biomarkers of interest (including but not limited to KL-6, surfactant protein D, CEA, CA125, CA19-9, CA15-3, CYFRA21-1, NSE, HE4, SCC, AFP, and CA242). Both controlled vocabulary (MeSH in PubMed, Emtree in EMBASE) and free-text terms were used to ensure comprehensive coverage. The full search strategy for PubMed is provided in Supplementary File 1. Inclusion and exclusion criteria The study selection followed the PICOS framework: Population (P): Patients diagnosed with rheumatoid arthritis (RA) based on the 1987 American College of Rheumatology (ACR) criteria or the 2010 ACR/European League Against Rheumatism (EULAR) classification criteria. Index test (I): Serum levels of tumor-associated biomarkers (including but not limited to CA125, CA19-9, CA15-3, CEA, KL-6, CYFRA21-1, NSE, HE4) measured using any standard immunoassay method (enzyme-linked immunosorbent assay [ELISA], chemiluminescence immunoassay [CLIA]). Comparator (C): RA patients without ILD (RA-non-ILD). ILD diagnosis in the RA-ILD group was required to be confirmed by high-resolution computed tomography (HRCT) of the chest, interpreted by a radiologist or pulmonologist according to established guidelines. Outcome (O): Primary outcomes were (1) pooled standardized mean difference (SMD) of biomarker concentrations between RA-ILD and RA-non-ILD groups, and (2) summary diagnostic accuracy metrics (area under the curve [AUC], sensitivity, specificity) for differentiating RA-ILD from RA-non-ILD. Study design (S): Observational studies, including case-control and cohort studies, were eligible. Exclusion criteria consisted of review articles, letters, case reports, commentaries, editorials, animal studies, and studies lacking sufficient quantitative data for meta-analysis. Data extraction and quality assessment Data extraction was independently performed by two investigators (Zeng Qiaoqian and Wu Qing) using a standardized, pre-piloted data extraction form. Discrepancies were resolved by discussion with a third reviewer (Wei Qijiao). Extracted data included: (1) study characteristics (first author, year, country, study design); (2) participant demographics (sample size, age, sex distribution, RA disease duration); (3) diagnostic criteria for RA and ILD; (4) tumor marker details (name, assay method, units); (5) quantitative data for the RA-ILD and RA-non-ILD groups (means, standard deviations, medians, interquartile ranges, or exact p-values for continuous outcomes); and (6) diagnostic accuracy data (sensitivity, specificity, cutoff values, AUC) where available. Methodological quality of included studies was assessed using the Newcastle-Ottawa Scale (NOS) for observational studies. The NOS evaluates three domains: selection of cohorts/cases, comparability, and ascertainment of outcome/exposure. Studies were categorized as high quality (NOS ≥ 7) or moderate quality (NOS = 5–6). The detailed NOS scores for each study are presented in Table 3. Table 3Results of quality assessment using the Newcastle-Ottawa ScaleNo.StudySelectionComparabilityOutcomeScoresRepresentativeness of the exposed cohortSelection of the non-exposed cohortAscertainment of exposureDemonstration that outcome of interest was not present at start of studyComparability of cohorts on the basis of the design or analysisAssessment of outcomeWas follow-up long enough for outcomes to occurAdequacy of follow up of cohorts1Zhang Xiaxia★★★★★★★☆☆72Song Xiao★★★★★★★☆☆73Wu Qi★★★★★★★☆☆74Zhang Jian★★★★★★★☆☆75Zhang Jintao★★★★★★★☆☆76Zhang Yuling★★★★★★★☆☆77Xia Junbo★★★★★★★☆☆79Mo Yuandong★★★★★★★☆☆79Xu Jiufei★★★★★★★☆☆710Ju Dandan★★★☆★★★☆☆611Li Linwei★★★★★★★☆☆712Chi Jing★★★★★★★☆☆713T. Wang★★★★★★★☆☆714Yan Qin★★★★★★★☆☆715Zheng Munan★★★★★★★☆☆716Go knan SARGIN★★★★★★★☆☆717Je´roˆme Avouac★★★★★★★☆☆718V colmenares★★★★★★★☆☆719Jia Xiguo★★★★★★★☆☆720Chen Gaofeng★★★★★★★☆☆722Wang Mengwei★★★★★★★☆☆723Jin Song★★★☆★★★☆☆624Wen Tao★★★★★★★☆☆725Zhou Junfei★★★★★★★☆☆726Liu Hongyan★★★★★★★☆☆727Jin Wan★★★★★★★☆☆728T.Lin★★★★★★★☆☆729Liu Liang★★★★★★★☆☆730Bo Liang★★★★★★★☆☆731yunyun Xu★★★★★★★☆☆732Wang Qingbao★★★★★★★☆☆733Rasha M. Hammoda★★★★★★★☆☆7NoStudySelectionComparabilityExposureScoresAdequate definition of casesRepresentativeness of the casesSelection of ControlsDefinition of controlsControl for important factorAscertainment of exposureSame method of ascertainment for cases and controlsNon-response rate21Wang Yingying★★★★★★★★☆834Tamao Nakashita★★★★★★★★☆835T Matsuo★★★★★★★★☆836Dina S. Fotoh★★★★★★★★☆8 Results of quality assessment using the Newcastle-Ottawa Scale Analysis methods Data synthesis was performed using Stata 11.0. As the primary outcome was continuous (biomarker concentrations), the pooled effect measure was the Standardized Mean Difference (SMD) with 95% confidence intervals (CI). For studies reporting data as medians with interquartile ranges, means and standard deviations were estimated using validated methods. A random-effects model (DerSimonian and Laird) was applied a priori for all meta-analyses, as it accounts for anticipated clinical and methodological heterogeneity across studies (e.g., differences in populations, assay methods, and study designs). Heterogeneity was assessed using Cochran’s Q test (p < 0.10 indicating significant heterogeneity) and the I² statistic. Subgroup analyses were pre-specified to explore potential sources of heterogeneity based on study design (cohort vs. non-cohort), geographic region (China vs. non-China), sample size (≥ 100 vs. <100), and mean age (≥ 60 vs. <60 years). These subgroups are presented in Table 4. Table 4Summary of the subgroup analysis resultsAnalysis N Random-effects modelFixed-effects modelHeterogeneitySMD (95% CI)P valueSMD (95% CI)P valueI2P valueCA125181.097 (0.474, 1.720)0.0010.716 (0.647, 0.784)0.0000.9870.000Subgroup1Cohort171.043 (0.401, 1.686)0.0010.668 (0.598, 0.737)0.0000.9870.000Non-cohort12.004 (1.642, 2.365)0.0002.004 (1.642, 2.365)0.0000.0000.000Subgroup2China171.190 (0.546, 1.833)0.0000.745 (0.676, 0.814)0.0000.9870.000Non-China1-0.456 (-0.892, -0.020)0.004-0.456 (-0.892, -0.020)0.0040.0000.000Subgroup3Sample ≥ 100130.625 (-0.074, 1.324)0.0800.695 (0.624, 0.765)0.0000.9890.000Sample < 10052.559 (0.601, 4.517)0.0101.027 (0.755, 1.300)0.0000.9790.000Subgroup4Age ≥ 60151.410 (0.907, 1.913)0.0001.253 (1.173, 1.332)0.0000.9720.000Age < 603-0.739 (-1.988, 0.511)0.247-0.835 (-0.970, -0.700)0.0000.9830.000CA199230.899 (0.564, 1.234)0.0000.461 (0.399, 0.522)0.0000.9620.000Subgroup1Cohort220.931 (0.580, 1.282)0.0000.463 (0.400, 0.525)0.0000.9640.000Non-cohort10.407 (0.098, 0.716)0.0100.407 (0.098, 0.716)0.0100.0000.000Subgroup2China220.911 (0.565, 1.257)0.0000.455 (0.394, 0.517)0.0000.9640.000Non-China10.725 (0.280, 1.170)0.0010.725 (0.280, 1.170)0.0000.0000.000Subgroup3Sample ≥ 100150.263 (-0.010, 0.537)0.0590.387 (0.322, 0.451)0.0010.9400.000Sample < 10082.707 (1.322, 4.091)0.0001.242 (1.034, 1.451)0.0000.9760.000Subgroup4Age ≥ 60190.935 (0.577, 1.294)0.0000.601 (0.531, 0.672)0.0000.9560.000Age < 6040.724 (-0.256, 1.704)0.1480.021 (-0.103, 0.146)0.7380.9730.000CEA220.576 (0.235, 0.917)0.0010.690 (0.604, 0.730)0.0000.9650.000Subgroup1Cohort210.540 (0.188, 0.892)0.0030.667 (0.360, 0.486)0.0000.9650.000Non-cohort11.330 (0.999, 1.662)0.0001.330 (0.999, 1.662)0.0000.0000.000Subgroup2China210.594 (0.242, 0.946)0.0010.701 (0.638, 0.763)0.0000.9660.000Non-China10.197 (-0.235, 0.629)0.3710.197 (-0.235, 0.629)0.3710.0000.000Subgroup3Sample ≥ 100150.541 (0.126, 0.957)0.0000.701 (0.635, 0.766)0.0000.9730.000Sample < 10070.656 (0.032, 1.280)0.0390.610 (0.427, 0.793)0.0000.9130.000Subgroup4Age ≥ 60180.565 (0.155, 0.974)0.0070.762 (0.690, 0.833)0.0000.9670.000Age < 6040.619 (-0.057, 1.294)0.0730.473 (0.349, 0.598)0.0000.9480.000KL-6151.678 (1.217, 2.139)0.0001.150 (1.059, 1.241)0.0000.9550.000Subgroup1Cohort121.717 (1.170, 2.264)0.0001.088 (0.991, 1.185)0.0000.9630.000Non-cohort31.563 (1.104, 2.021)0.0001.594 (1.334, 1.855)0.0000.6470.000Subgroup2China101.156 (0.798, 1.514)0.0000.957 (0.858, 1.055)0.0000.9080.000Non-China52.751 (1.422, 4.080)0.0002.219 (1.987, 2.452)0.0000.9670.000Subgroup3Sample ≥ 10081.555 (0.981, 2.128)0.0001.065 (0.964, 1.165)0.0000.9640.000Sample < 10071.832 (0.940, 2.724)0.0001.526 (1.315, 1.737)0.0000.9420.000Subgroup4Age ≥ 60101.460 (0.854, 2.066)0.0001.057 (0.932, 1.183)0.0000.9650.000Age < 6052.146 (1.214, 3.078)0.0001.253 (1.121, 1.386)0.0000.9540.000AFP70.162 (0.082, 0.241)0.9190.162 (0.082, 0.241)0.0000.9160.000Subgroup1Cohort70.162 (0.082, 0.241)0.9190.162 (0.082, 0.241)0.0000.9160.000Non-cohort0NANANANANANASubgroup2China70.162 (0.082, 0.241)0.9190.162 (0.082, 0.241)0.0000.9160.000Non-China0NANANANANANASubgroup3Sample ≥ 10050.092 (-0.26, 0.444)0.6090.188 (0.106, 0.271)0.0000.9370.000Sample < 100200.213 (-0.641, 0.214)0.328-0.209 (-0.514, 0.096)0.180.4890.162Subgroup4Age ≥ 6050.055 (-0.315, 0.425)0.7710.303 (0.207, 0.399)0.0000.9110.000Age < 602-0.140 (-0.28, 0.000)0.0510.162 (0.082, 0.241)0.0510.0000.492CA153121.698 (1.162, 2.234)0.0000.664 (0.571, 0.757)0.0000.9620.000Subgroup1Cohort121.698 (1.162, 2.234)0.0000.664 (0.571, 0.757)0.0000.9620.000Non-cohort0NANANANANANASubgroup2China101.918 (1.302, 2.534)0.0000.657 (0.561, 0.753)0.0000.9690.000Non-China20.768 (0.391, 1.144)0.0000.768 (0.391, 1.144)0.0000.0000.564Subgroup3Sample ≥ 10070.637 (0.412, 0.862)0.0000.564 (0.466, 0.662)0.0000.7710.000Sample < 10054.310 (1.820, 6.801)0.0011.626 (1.323, 1.930)0.0000.9820.000Subgroup4Age ≥ 6081.634 (0.95, 2.319)0.0000.830 (0.693, 0.967)0.0000.9530.000Age < 6042.070 (0.944, 3.195)0.0000.522 (0.395, 0.649)0.0000.9770.000 Summary of the subgroup analysis results Sensitivity analysis was conducted using the leave-one-out method, in which the meta-analysis was iteratively repeated after excluding one study at a time, to assess the robustness of the pooled estimates. Publication bias was evaluated by visual inspection of funnel plots and by Egger’s linear regression test. This analysis was performed only for outcomes where a sufficient number of studies (≥ 10) were available. When significant publication bias was detected (p < 0.05), the trim-and-fill method was applied to adjust for the potential bias and provide corrected effect estimates. Synthesis of reported diagnostic accuracy data We extracted diagnostic accuracy metrics (AUC, optimal cutoff, sensitivity, specificity) from studies that reported these data for distinguishing RA-ILD from RA-non-ILD. These data were tabulated and summarized descriptively. This analysis does not constitute a formal diagnostic test accuracy meta-analysis; no pooled sensitivity, specificity, likelihood ratios, or hierarchical models (e.g., bivariate or HSROC) were applied. The reported AUC values and related metrics are presented as originally reported by each study, serving as an exploratory summary rather than a pooled estimate of diagnostic performance. Results are presented in the results section and Fig. 5. ROC curve graph for CA125, CA153, CA199, CEA, CYFRA21-1 and KL-6 Results Study selection and characteristics A total of 2,696 potentially relevant articles were initially identified through systematic literature searches employing predefined search strategies and eligibility criteria. After removing 79 duplicate records, 2,617 publications underwent title/abstract screening. This preliminary screening excluded 279 review articles, 190 case reports, 131 book chapters/commentaries, 55 conference abstracts, 27 animal studies, and 1,082 articles focused on diagnostic/therapeutic approaches. Subsequently, 783 articles were excluded due to irrelevance to the study’s scope: 379 unrelated to RA, 203 unrelated to ILD, and 201 unrelated to tumor biomarkers. Full-text evaluation was conducted for 70 potentially eligible articles. After detailed assessment, we excluded 5 publications with overlapping datasets, 2 studies lacking tumor marker classification, 26 articles without ILD/non-ILD group comparisons, and 1 study failing to report quantitative tumor marker data. Ultimately, 36 studies [5–39] met all inclusion criteria and were selected for final analysis (Fig. 1). Flow diagram of the literature search process Study characteristics All included studies underwent rigorous full-text evaluation, with key study characteristics systematically documented (Tables 1 and 2). Data extraction was performed across 32 retrospective studies, 2 case-control studies, 2 cross-sectional studies. The final analytical cohort comprised 4,958 RA patients without ILD and 2,342 RA patients with ILD, aggregating data from 7,300 individuals. Demographic parameters including mean age, gender distribution, and study period were comprehensively recorded in Table 1, stratified by ILD status. Table 1Characteristics of the studies included in the meta-analysisNO.First authorYearCountryJournalStudy designSample sizeSample sexSample ageSurvey yearRA-nonILDRA-ILDRA-nonILDF/MRA-ILDF/MRA-nonILDRA-ILD1Zhang Xiaxia2021ChinaThesis of Master’s Degree in Xinjiang Medical UniversityRetrospective cohort study858455/2655/2962.8 ± 10.365.3 ± 9.02014.10-2019.122Song Xiao2021ChinaThesis of Master’s Degree in Shandong University of Traditional Chinese MedicineRetrospective cohort study704643/2729/1757.3 ± 12.664.6 ± 9.22016.1-2017.123Wu Qi2021ChinaThesis of Master’s Degree in Shantou UniversityRetrospective cohort study455282392/63201/8138.4 ± 12.347.2 ± 13.42015.1-2020.74Zhang Jian2018ChinaChina prac MedRetrospective cohort study404028/1229/1136.0 ± 17.535.0 ± 16.52015.1-2020.75Zhang Jintao2018ChinaChina Journal of Modern MedicineRetrospective cohort study17850147/3130/2058.0 (49.0, 63.0)61.0 (57.8, 69.0)2013.2-2017.26Zhang Yuling2017ChinaThesis of Master’s Degree in Tianjin Medical UniversityRetrospective cohort study423933/926/1357.8 ± 10.762.0 ± 7.82015.4-2017.17Xia Junbo2012ChinaJournal of Third Military Medical UniversityRetrospective cohort study443826/1822/1670.2 ± 14.368.2 ± 10.32008.1-2011.18Mo Yuandong2017ChinaThesis of Master’s Degree in Guangxi Medical UniversityRetrospective cohort study949476/1860/3454.8 ± 11.659.0 ± 11.82006.12-2012.129Xu Jiufei2018ChinaChina Modern DoctorRetrospective cohort study22660140/8628/3257.3 ± 9.866.1 ± 8.02015.1-2018.1210Ju Danna2018ChinaThesis of Master’s Degree in Qingdao UniversityRetrospective cohort study3321NANANANA2015.12-2017.1211Li Linwei2015ChinaChinese Journal of Respiratory and Critical CareRetrospective cohort study342512/2214/1157.3 ± 12.463.6 ± 11.92011.5-2014.1112Chi Jing2013ChinaAdvances in Modern BiomedicineRetrospective cohort study533532/2120/1562.4 ± 3.964.1 ± 4.02011.3-2013.313T. Wang2016ChinaClinical and experimental medicineRetrospective cohort study832853/3012/1654.1 ± 15.563.4 ± 11.32011–201414Yan Qin2022ChinaArthritis research &TherapyRetrospective cohort study787558/2045/3045.2 ± 13.362.8 ± 8.72020.2-2021.1115Munan Zheng2021ChinaRheumatol TherRetrospective cohort study262420/610/1456.6 ± 10.362.9 ± 11.82019.1-2019.916Go knan SARGIN2018TurkeyArch RheumatolRetrospective cohort study404333/730/1358.5 ± 12.760.1 ± 11.52014.1-2017.1017Je´roˆme Avouac2020France/Japan/SwitzerlandPLOS ONERetrospective cohort study1074080/2722/1862.0 ± 12.071.0 ± 15.0NA18V colmenares2024SwedenScand J RheumatolRetrospective Cohort study441028/166/458.5 ± 13.771.8 ± 6.7NA19Jia Xiguo2025ChinaBMC Pulmonary medicineRetrospective cohort study13511533/10234/8154.9 ± 10.962.5 ± 9.52018.1-2020.1220Chen Gaofeng2021ChinaThesis of Master’s Degree in Anhui Medical UniversityRetrospective cohort study11511567/4867/4859.3 ± 9.860.0 ± 110.62015.1-2019.221Wang Yingying2024ChinaThesis of Master’s Degree in Yan’an UniversityCross-sectional Study15955124/3525/3056.0 (47.0, 64.0)64.0 (59.0, 69.0)2020.10-2023.1022Wang Mengwei2024ChinaThesis of Master’s Degree in Wanna Medical UniversityRetrospective cohort study17670147/2939/3157.7 ± 13.364.2 ± 9.52021.4-2023.1223Jin Song2018ChinaGenomics and Applied BiologyRetrospective cohort study6030NANANANANA24Wen Tao2023ChinaModern Medical imagelogyRetrospective cohort study305010/2014/3655.4 ± 12.755.1 ± 12.92018.12-2020.1225Zhou Junfei2016ChinaThesis of Master’s Degree in Zheng Zhou UniversityRetrospective cohort study515133/1829/2257.8 ± 13.059.8 ± 10.32015.8-2016.326Liu Hongyan2024ChinaContemporary MedicineRetrospective cohort study515120411/10490/3060 0.0(50.0, 67.0)65 0.0(56.3, 72.0)2018.1-2020.1027Jin Wan2024ChinaClinical RheumatologyRetrospective cohort study497634/1556/2057.7 ± 13.765.7 ± 10.72019.10-2023.1028T.Lin2022ChinaClinical and Experimental RheumatologyRetrospective cohort study564629/2731/1558.3 ± 8.863.4 ± 8.2NA29Liu Liang2021ChinaFrontiers in medicineRetrospective cohort study12048111/936/1262 0.0(53.0, 70.0)67.0 (57.0,73.0)2020.1-2020.1030Bo Liang2024ChinaImmunological Investigations A Journal of Molecular and Cellular ImmunologyRetrospective cohort study453831/1422/1657.6 ± 12.465.2 ± 9.92022.11-2023.1031Xu Yunyun2022ChinaClinical RheumatologyRetrospective cohort study1215284941/274179/10558.0 ± 12.069.0 ± 9.42016.08-2020.1032Wang Qingbao2024ChinaChina Journal of Modern MedicineRetrospective cohort study703167/330/152.9 ± 7.656.3 ± 8.92021.1-2023.433Rasha M. Hammoda2021EgyptThe Egyptian Journal of ImmunologyRetrospective cohort study256418/749/1543.9 ± 9.647.7 ± 11.8NA34Tamao Nakashita2014JapanBMJ OpenCase–control study441422/229/565.8 ± 8.867.0 ± 9.62006.4-2012.335T Matsuo2018JapanScandinavian Journal of RheumatologyCross-sectional study28626250/3621/562.9 ± 12.969.8 ± 9.52011.5.1-2012.11.136Dina S. Fotoh2021EgyptClinical RheumatologyCase–control study757536/3935/4045.3 ± 11.547.4 ± 10.52019.1-2020.1Note: RA: Rheumatoid Arthritis; ILD: Interstitial Lung Disease; F: Female; M: Male Characteristics of the studies included in the meta-analysis Note: RA: Rheumatoid Arthritis; ILD: Interstitial Lung Disease; F: Female; M: Male Table 2Summary of data available for meta-analysisNo.AuthorTumor markersTest methodDiagnostic methods for RA-ILDDetected dataχ2/t testP valueRA-nonILDRA-ILD1Zhang XiaxiaCEA (ng/mL)ChemiluminescenceImmunoassay(CLIA)1987 ACR or 2010 ACR/EULAR diagnostic criteria for RA2002 American Thoracic Society (ATS) and European Respiratory Society (ERS) diagnostic criteria for Idiopathic Pulmonary Fibrosis (IPF)6.6 (3.0, 10.5)3.2 (1.8, 5.4)-3.10.0CA125 (U/mL)1.6 (1.0, 2.0)32.9(16.6, 82.0)-11.10.0CA153 (U/mL)12.1 (8.2, 19.4)22.0 (16.0, 36.3)-5.70.0CA199 (U/mL)8.8 (6.4, 13.5)17.3 (6.5, 39.7)-3.70.0CYFRA21-1 (ng/mL)2.0 (1.5, 2.8)3.3 (2.3, 4.9)-5.50.0NSE (ng/mL)14.5 (12.0, 16.8)15.0 (12.5, 19.7)-2.00.0SCC (ng/mL)0.6 (0.5, 0.8)0.9 (0.7, 1.5)-5.10.02Song XiaoAFP (ng/mL)NA2010 ACR/EULAR diagnostic criteria for RA2015 ATS/ERS diagnostic criteria for IPF3.0 ± 1.73.2 ± 1.7NA0.6CEA (ng/mL)2.1 ± 2.64.3 ± 4.1NA0.0CA199 (U/mL)9.9 ± 11.044.6 ± 64.6NA0.0CA724 (U/mL)4.3 ± 5.75.1 ± 6.6NA0.13Wu QiKL-6 (U/mL)NA2010 ACR/EULAR diagnostic criteria for RA2013 ATS/ERS diagnostic criteria for IPF175.1 (129.0, 244.8)310.6 (155.0, 327.0)-3.40.0CA125 (U/mL)14.0 (10.4, 20.6)14.7 (10.5, 21.2)-1.00.3CA153 (U/mL)7.6 (5.7, 10.6)9.5 (7.0, 13.4)-5.60.0CA199 (U/mL)3.7 (2.0, 8.7)3.9 (2.0, 8.1)-0.10.9AFP (ng/mL)2.1 (1.5, 2.8)2.0 (1.6, 3.2)0.80.4CEA (ng/mL)1.2 (0.8, 1.7)1.7 (1.0, 3.5)-6.30.04Zhang JianCEA (ng/mL)CLIAACR/EULAR diagnostic criteria for RAFor RA-ILD not available2.4 ± 1.06.3 ± 2.78.6< 0.05CA153 (U/mL)10.1 ± 4.796.7 ± 12.042.4< 0.05CA199 (U/mL)158.0 ± 90.8320.0 ± 110.231.7< 0.055Zhang JintaoCEA (ng/mL)NA1987 ACR/EULAR diagnostic criteria for RA2002 ATS/ERS diagnostic criteria for IPF1.4 (0.8, 2.6)2.5 (1.5, 4.3)-4.80.0CA125 (U/mL)11.1 (7.9, 14.3)13.7 (9.6, 36.0)-3.20.0CA199 (U/mL)6.0 (3.8, 10.0)10.8 (6.5, 23.3)-4.50.06Zhang YulingKL-6 (U/mL)CLIA2010 ACR/EULAR diagnostic criteria for RA2002 ATS/ERS diagnostic criteria for IPF76.9 ± 80.3269.7 ± 210.2NA< 0.05AFP (ng/mL)2.4 ± 1.12.0 ± 1.1NA0.2CEA (ng/mL)2.3 ± 1.43.1 ± 2.4NA0.1CA199 (U/mL)11.3 ± 10.921.1 ± 22.9NA0.0CA242 (U/mL)6.9 ± 5.59.0 ± 10.1NA0.3NSE (ng/mL)13.3 ± 4.211.4 ± 6.3NA0.3CYFRA21-1 (ng/mL)2.3 ± 1.12.2 ± 0.8NA0.7SCC (ng/mL)0.6 ± 0.50.9 ± 1.0NA0.37Xia JunboCEA (ng/mL)CLIA1987 ACR/EULAR diagnostic criteria for RAHRCT for ILD2.2 ± 1.06.2 ± 2.2NA< 0.01CA153 (U/mL)10.5 ± 4.496.1 ± 12.2NA< 0.01CA199 (U/mL)8.8 ± 3.672.3 ± 12.4NA< 0.018Mo YuandongCEA (ng/mL)NA1987 ACR or 2010 ACR/EULAR diagnostic criteria for RA2000 ATS/ERS diagnostic criteria for IPF1.4 ± 1.42.2 ± 2.1NA0.0CA125 (U/mL)12.1 ± 9.620.8 ± 21.3NA0.0CA153 (U/mL)10.6 ± 8.414.8 ± 10.8NA0.0CA199 (U/mL)4.8 ± 7.27.3 ± 15.3NA0.29Xu jiufeiAFP (ng/mL)CLIA2010 ACR/EULAR diagnostic criteria for RA2013 ATS/ERS diagnostic criteria for IPF2.0 (1.4, 2.5)2.2 (1.9, 2.7)1.20.1CA125 (U/mL)19.7 (12.0, 32.9)42.4 (24.5, 49.7)2.8< 0.001CEA (ng/mL)2.4 (1.6, 3.7)3.0 (1.6, 5.9)2.00.001CA199 (U/mL)6 0.6 (3.2, 16.3)21.1 (3.5, 67.9)2.3< 0.00110Ju DannaCA125 (U/mL)CLIA2010 ACR/EULAR diagnostic criteria for RA2002 ATS/ERS diagnostic criteria for IPF10.0 (6.7, 13.4)21.5 (15.1, 31.8)3.9< 0.001CA199 (U/mL)6 0.8 (1.9, 26.3)13.1 (2.3, 23.3)1.00.311Li LinweiCEA (ng/mL)NA1987 ACR/EULAR diagnostic criteria for RAHRCT for ILD1.9 ± 0.72.8 ± 2.1NA0.1CA153 (U/mL)11.7 ± 2.724.7 ± 13.3NA0.0CA199 (U/mL)12.0 ± 9.349.7 ± 89.4NA0.1CA125 (U/mL)26.5 ± 10.568.7 ± 57.5NA0.012Chi JingCEA (ng/mL)Protein Chip System for Multi-tumor Marker Detection1987 ACR/EULAR diagnostic criteria for RAHRCT for ILD3.3 ± 0.97.1 ± 1.32.40.0CA125 (U/mL)13.7 ± 6.497.3 ± 14.62.40.0CA199 (U/mL)10.6 ± 4.388.2 ± 13.82.40.013T. WangCEA (ng/mL)CLIA1987 ACR/EULAR diagnostic criteria for RAHRCT for ILD1.6 (0.2, 5.7)2.1 (0.6, 11.4)NA0.1CA125 (U/mL)21.2 (4.6, 205.6)38.3 (12.6, 199.8)NA0.0CA153 (U/mL)13.0 (4.9, 47.6)15.8 (9.0, 94.0)NA0.0CA199 (U/mL)8.0 (0.6, 153.1)13.9 (0.6, 355.0)NA0.014Yan QinCEA (ng/mL)CLIA1987 ACR/EULAR diagnostic criteria for RAHRCT for ILD1.4 (1.0, 2.0)2.3 (1.2, 3.8)NA0.0CA199 (U/mL)5.0 (3.1, 8.0)9.1 (5.6, 22.4)NA0.0CA242 (U/mL)3.9 (2.9, 6.0)6.9 (4.0, 13.1)NA0.0KL-6 (U/mL)260.8 (188.1, 368.8)470.5 (280.9, 804.9)NA0.015Munan ZhengCA125 (U/mL)CLIA1987 ACR/EULAR diagnostic criteria for RAHRCT for ILD9.0 (3.5, 17.8)40.0 (7.1, 75.0)NA0.0CA199 (U/mL)16.2 (6.7, 42.8)56.8 (11.8, 521.6)NA0.0CEA (ng/mL)0.9 (0.1, 2.8)1.6 (0.5, 3.8)NA0.0KL-6 (U/mL)227.0 (136.0, 581.0)558.0 (221.0, 3575.0)NA0.016Go knan SARGINCEA (ng/mL)NA2010 ACR/EULAR diagnostic criteria for RAHRCT for ILD2.7 ± 1.82.9 ± 1.6NA0.1CA153 (U/mL)14.7 (8.3, 19.6)20.4 (15.2, 35.1)NA0.0CA125 (U/mL)12.8 (9.2, 21.8)18.1 (11.5, 26.8)NA0.0CA199 (U/mL)7.3 (3.7, 11.9)11.9 (5.6, 29.5)NA0.017Je´roˆme AvouacKL-6 (U/mL)ELISA1987 ACR/EULAR diagnostic criteria for RAHRCT for ILD376.0 ± 26.0961.0 ± 128.0NA0.018V colmenaresCA153 (pg/mL)ELISAACR/EULAR diagnostic criteria for RAHRCT for ILD30.0 ± 22.551.1 ± 21.5NANAYKL-40 (ng/mL)31.7 ± 35.260.5 ± 73.0NANA19Jia XiguoCA125 (U/mL)CLIA1987 ACR or 2010 ACR/EULAR diagnostic criteria for RAHRCT for ILD9.1 (6.7, 13.0)10.7 (7.0, 17.3)NA0.1CA153 (U/mL)6.4 (4.7, 10.5)10.4 (6.1, 17.2)NA0.0CA199 (U/mL)14.2 (8.7, 21.0)17.5 (10.6, 28.0)NA0.0CA242 (IU/mL)3.7 (2.6, 5.9)4.3 (3.0, 6.6)NA0.1CA50 (U/mL)4.8 (2.8, 7.7)6.5 (4.1, 10.7)NA0.0CA724 (U/mL)1.5 (0.9, 3.2)1.6 (0.9, 3.5)NA0.1CEA (ng/mL)1.0 (0.5, 1.5)1.3 (0.6, 2.0)NA0.0CYFRA21-1 (ng/mL)2.1 (1.5, 3.0)2.7 (1.9, 4.3)NA0.0NSE (ng/mL)3.1 (2.6, 4.0)3.6 (2.8, 4.6)NA0.1Pro GRP (pg/mL)31.0 (23.3, 43.7)34.1 (25.2, 45.2)NA0.1SCC (ng/mL)1.0 (0.7, 1.6)1.1 (0.7, 1.5)NA0.1KL-6 (U/mL)229.0 (163.0, 311.0)287.0 (167.0, 531.0)NA0.120Chen GaofengKL-6 (U/mL)CLIA1987 ACR/EULAR diagnostic criteria for RAHRCT for ILD187.0 (154.0, 258.0)466.0 (281.0, 717.0)NA0.021Wang YingyingCEA (ng/mL)NA1987 ACR/EULAR diagnostic criteria for RAHRCT for ILD2.4 (1.7, 3.3)4.4 (2.2, 6.7)-3.60.0CA125 (U/mL)13.4 (8.1, 20.5)59.7 (23.1, 85.7)-4.70.0CA199 (U/mL)7.1 (3.1, 8.6)19.3 (8.2, 29.4)-4.20.022Wang MengweiCEA (ng/mL)NA2010 ACR/EULAR diagnostic criteria for RA2018 ATS/ERS diagnostic criteria for IPF2.9 ± 2.03.9 ± 2.92.60.0CA125 (U/mL)11.9 ± 7.348.1 ± 69.54.30.0CA199 (U/mL)9.3 ± 6.930.4 ± 60.22.90.0CA153 (U/mL)12.5 ± 8.410.1 ± 5.5-2.60.023Jin SongKL-6 (U/mL)ELISANA210.5 ± 67.4419.8 ± 183.07.90.024Wen TaoKL-6 (U/mL)ELISA1987 ACR or 2010 ACR/EULAR diagnostic criteria for RAHRCT for ILD88.5 ± 20.3286.4 ± 50.620.40.025Zhou JunfeiKL-6 (U/mL)NA1987ACR/EULAR diagnostic criteria for RA2010 Bongartz T criteria for RA-ILD224.1 ± 52.9419.8 ± 322.6NA0.026Liu HongyanCEA (ng/mL)NA1987 ACR/EULAR diagnostic criteria for RAHRCT for ILD1.8 (1.1, 2.7)2.3 (1.5, 3.1)-3.80.0AFP (ng/mL)2.4 (1.8, 3.1)2.4 (1.9, 3.0)-0.10.9CA125 (U/mL)14.4 (9.4, 2.7)16.5 (10.5, 28.3)-2.70.0CA199 (U/mL)7.3 (4.2, 11.8)10.3 (6.3, 18.0)-4.80.027Jin WanCA125 (U/mL)NA2010ACR/EULAR diagnostic criteria for RAHRCT for ILD11.8 (7.7, 16.2)18.0 (12.9, 27.3)NA0.0CA199 (U/mL)20.0 (10.6, 40.4)35.2 (23.9, 76.9)NA0.0NSE (ng/mL)13.1 (10.8, 15.9)14.0 (12.3, 15.9)NA0.1CEA (ng/mL)1.1 (0.7, 1.4)1.2 (0.8, 2.0)NA0.1CYFRA21-1 (ng/mL)2.1 (1.6, 2.4)2.5 (1.8, 3.4)NA0.0Pro GRP (pg/mL)41.5 (27.5, 52.2)44.1 (34.1, 62.4)NA0.228T.LinHE4 (pg/mL)CLIA2010 ACR/EULAR diagnostic criteria for RAHRCT for ILD82.7 ± 26.2141.8 ± 65.9NA0.029Liu LiangHE4 (pg/mL)CLIA2010 ACR/EULAR diagnostic criteria for RAFor RA-ILD not available108.3 ± 13.8169.9 ± 22.7-2.20.030Bo LiangKL-6 (U/mL)ELISA2010 ACR/EULAR diagnostic criteria for RAHRCT for ILD271.7 ± 132.9513.9 ± 266.0NA0.0YKL-40 (ng/mL)34.8 ± 22.257.7 ± 33.4NA0.031yunyun XuCEA (ng/mL)NA2010 ACR/EULAR diagnostic criteria for RAHRCT for ILD2.4 ± 5.03.9 ± 3.3NA0.0AFP (ng/mL)2.6 ± 1.62.4 ± 1.2NA0.1CA125 (U/mL)18.2 ± 21.547.8 ± 46.9NA0.0CA199 (U/mL)10.9 ± 12.232.1 ± 47.7NA0.0CYFRA21-1 (ng/mL)2.5 ± 3.43.4 ± 1.8NA0.032Wang QingbaoAFP (ng/mL)NAACR/EULAR diagnostic criteria for RAHRCT for ILD2.5 ± 1.12.4 ± 1.20.20.8CEA (ng/mL)1.5 (1.1, 2.4)1.9 (1.3, 3.2)1.60.1CA199 (U/mL)2.7 (2.0, 4.7)3.4 (2.0, 8.7)1.50.1CA153 (U/mL)8.6 (6.7, 11.6)9.8 (7.0, 14.5)1.50.1CA211 (U/mL)1.2 (0.9, 1.7)1.39 (1.1, 1.9)1.50.1CA125 (U/mL)12.4 (9.2, 16.0)16.6 (10.9, 22.7)2.80.0HE4 (pmol/mL)37.5 (31.9, 43.1)48.9 (43.0, 66.4)5.00.033Rasha M. HammodaKL-6 (U/mL)enzyme-linked immunosorbent assay kit2010 ACR/EULAR diagnostic criteria for RAHRCT for ILD324.0(281.0, 418.0)1235.0 (1100.0, 1455.0)NA0.034Tamao NakashitaKL-6 (U/mL)NADiagnostic criteria for RA not availableHRCT for ILD388.0 ± 347.0339.0 ± 172.0NA1.035T MatsuoKL-6 (U/mL)ELISAACR/EULAR diagnostic criteria for RAHRCT for ILD278.5 ± 147.9471.3 ± 326.8NA0.036Dina S. FotohKL-6 (U/mL)latex-enhanced immunoturbidimetric assay2010 ACR/EULAR diagnostic criteria for RAHRCT for ILD237.5 ± 51.91025.5 ± 419.6NA0.0 Summary of data available for meta-analysis Chemiluminescence Immunoassay (CLIA) 1987 ACR or 2010 ACR/EULAR diagnostic criteria for RA 2002 American Thoracic Society (ATS) and European Respiratory Society (ERS) diagnostic criteria for Idiopathic Pulmonary Fibrosis (IPF) 2010 ACR/EULAR diagnostic criteria for RA 2015 ATS/ERS diagnostic criteria for IPF 2010 ACR/EULAR diagnostic criteria for RA 2013 ATS/ERS diagnostic criteria for IPF ACR/EULAR diagnostic criteria for RA For RA-ILD not available 1987 ACR/EULAR diagnostic criteria for RA 2002 ATS/ERS diagnostic criteria for IPF 2010 ACR/EULAR diagnostic criteria for RA 2002 ATS/ERS diagnostic criteria for IPF 1987 ACR/EULAR diagnostic criteria for RA HRCT for ILD 1987 ACR or 2010 ACR/EULAR diagnostic criteria for RA 2000 ATS/ERS diagnostic criteria for IPF 2010 ACR/EULAR diagnostic criteria for RA 2013 ATS/ERS diagnostic criteria for IPF 2010 ACR/EULAR diagnostic criteria for RA 2002 ATS/ERS diagnostic criteria for IPF 1987 ACR/EULAR diagnostic criteria for RA HRCT for ILD 1987 ACR/EULAR diagnostic criteria for RA HRCT for ILD 1987 ACR/EULAR diagnostic criteria for RA HRCT for ILD 1987 ACR/EULAR diagnostic criteria for RA HRCT for ILD 1987 ACR/EULAR diagnostic criteria for RA HRCT for ILD 2010 ACR/EULAR diagnostic criteria for RA HRCT for ILD 1987 ACR/EULAR diagnostic criteria for RA HRCT for ILD ACR/EULAR diagnostic criteria for RA HRCT for ILD 1987 ACR or 2010 ACR/EULAR diagnostic criteria for RA HRCT for ILD 1987 ACR/EULAR diagnostic criteria for RA HRCT for ILD 1987 ACR/EULAR diagnostic criteria for RA HRCT for ILD 2010 ACR/EULAR diagnostic criteria for RA 2018 ATS/ERS diagnostic criteria for IPF 1987 ACR or 2010 ACR/EULAR diagnostic criteria for RA HRCT for ILD 1987ACR/EULAR diagnostic criteria for RA 2010 Bongartz T criteria for RA-ILD 1987 ACR/EULAR diagnostic criteria for RA HRCT for ILD 2010ACR/EULAR diagnostic criteria for RA HRCT for ILD 2010 ACR/EULAR diagnostic criteria for RA HRCT for ILD 2010 ACR/EULAR diagnostic criteria for RA For RA-ILD not available 2010 ACR/EULAR diagnostic criteria for RA HRCT for ILD 2010 ACR/EULAR diagnostic criteria for RA HRCT for ILD ACR/EULAR diagnostic criteria for RA HRCT for ILD 2010 ACR/EULAR diagnostic criteria for RA HRCT for ILD Diagnostic criteria for RA not available HRCT for ILD ACR/EULAR diagnostic criteria for RA HRCT for ILD 2010 ACR/EULAR diagnostic criteria for RA HRCT for ILD Quality assessment Methodological quality was assessed using the Newcastle-Ottawa Scale (NOS). Of the 36 included studies, 28 studies (77.8%) were rated as high quality (NOS score ≥ 7), and 8 studies (22.2%) were rated as moderate quality (NOS score 5–6). No studies were considered low quality (NOS < 5) and were excluded. The detailed NOS scores for each study are presented in Table 3. Analysis of methodological weaknesses revealed several common issues. First, regarding selection, several studies did not explicitly state that the outcome of interest (ILD) was not present at the start of the study, which could introduce selection bias. Second, regarding comparability, only a subset of studies controlled for key confounders such as age, sex, smoking status, or RA disease duration, potentially affecting the validity of the observed associations. Third, regarding outcome, while all studies used HRCT to diagnose ILD, the assessment was not always blinded to the patient’s RA and biomarker status, and the adequacy of follow-up for the cohort studies was often not reported. These methodological considerations highlight the need for cautious interpretation of the pooled estimates. Data synthesis The distribution of biomarkers across the 36 eligible studies is summarized in Table 2. The analysis revealed distinct patterns of biomarker elevation between RA-ILD and RA-non-ILD groups. Significantly elevated levels were observed for CA125 (SMD = 1.097, 95%CI [0.474–1.720], P = 0.001), CA199 (SMD = 0.899, 95%CI [0.564–1.234], P < 0.001), CA153 (SMD = 1.698, 95%CI [1.162–2.234], P < 0.001), CEA (SMD = 0.576, 95%CI [0.235–0.917], P = 0.001), KL-6 (SMD = 1.678, 95%CI [1.217–2.139], P < 0.001), NSE (SMD = 0.320, 95%CI [0.162–0.479], P < 0.001), and HE4 (SMD = 1.869, 95%CI [1.603–2.135], P < 0.001) in the RA-ILD cohort. Notably, KL-6 demonstrated the strongest association among all biomarkers. In contrast, no statistically significant differences were detected for AFP, CA242, or SCC between the groups. After adjusting for publication bias using the trim-and-fill method, CYFRA21-1 was also significantly elevated in RA-ILD patients (SMD = 0.475, 95%CI [0.058–0.892], p = 0.026), consistent with the findings reported in the abstract. To aid clinical interpretation, the magnitude of the effect sizes was assessed using standard thresholds. HE4 and CA153 showed large effect sizes (SMD > 0.8), with HE4 demonstrating a massive effect (SMD = 1.87), indicating a very strong separation between RA-ILD and RA-non-ILD groups. KL-6 and CA125 also demonstrated large effect sizes (SMD = 1.68 and 1.10, respectively). CA199 and CEA showed moderate-to-large effects (SMD = 0.90 and 0.58, respectively). In contrast, NSE showed a small but statistically significant effect (SMD = 0.32). Both studies investigating YKL-40 reported significant elevations in RA-ILD patients, while the two studies examining Pro GRP and CA724 found no significant associations. Limited evidence from single studies suggested non-significant differences for CA211 but significant elevation of CA50 in RA-ILD. Note that several biomarkers (YKL-40, Pro GRP, CA724, CA211, and CA50) were investigated in fewer than three studies, indicating the need for further validation. The complete comparative analysis is visually presented in Fig. 2. Forest plots for CA125, CA153, CA199, CEA, CYFRA21-1, KL-6, NSE and HE4 Heterogeneity analysis Significant heterogeneity was observed across studies for CA125, CA199, CEA, KL-6, AFP, and CA153. To explore its sources, we conducted pre-specified subgroup analyses based on study design, geographic region, sample size, and mean age (Table 4). The subgroup analyses revealed several patterns. For CA125, heterogeneity remained high across all subgroups, though the effect size appeared larger in studies with sample size < 100 (SMD = 2.56) and in Chinese cohorts (SMD = 1.19). For CA199, heterogeneity was markedly reduced in studies with sample size ≥ 100 (I²=94.0%) compared to those with sample size < 100 (I²=97.6%), suggesting that smaller studies contributed substantially to the overall heterogeneity. Similarly, for CEA and KL-6, studies conducted in non-Chinese populations showed lower heterogeneity but with wider confidence intervals, likely due to the limited number of studies (n = 1 for CEA, n = 5 for KL-6). For AFP, no significant difference was observed between RA-ILD and RA-non-ILD groups in any subgroup, consistent with the main analysis. For CA153, heterogeneity was notably lower in studies with sample size ≥ 100 (I²=77.1%) compared to those with sample size < 100 (I²=98.2%), and in non-Chinese populations (I²=0.0%), though the latter was based on only two studies. Overall, while the subgroup analyses did not identify a single predominant source of heterogeneity, they suggest that sample size and geographic region may contribute to the observed variability, particularly for CA199 and CA153. The persistent heterogeneity across subgroups may also reflect unmeasured factors, including differences in assay platforms(e.g., ELISA vs. chemiluminescence), variations in RA disease activity levels across cohorts, patient selection criteria, or unadjusted confounders. Sensitivity analysis Sensitivity analysis was performed using the leave-one-out method to assess the robustness of the pooled estimates for key biomarkers. For CA125, CA153, and CA199, the pooled estimate fell outside the 95% confidence interval after excluding the study by Wu Qi et al., indicating that this study had a notable influence on the overall effect for these markers. However, the direction and significance of the effect remained unchanged for all three biomarkers after removal of any single study. For KL-6, the pooled SMD remained statistically significant (range: 1.48 to 1.79) after the sequential removal of each study, with no single study substantially altering the overall estimate or its significance, indicating that the strong association between KL-6 and RA-ILD was highly robust. For HE4, the pooled SMD remained statistically significant and stable (range: 1.81 to 1.93) after the sequential removal of each of the three included studies, confirming the robustness of this finding. The results of the sensitivity analysis are shown in Fig. 3 (for CA125, CA153, and CA199) and are summarized in the text for KL-6 and HE4. Sensitivity analysis for CA125, CA153, and CA199 Publication bias The results from the Egger test showed that there was no potential bias in SCC (P = 0.69), NSE (P = 0.18), CEA (P = 0.86), CA242 (P = 0.83), CA199 (P = 0.18), CA153 (P = 0.46), CA125 (P = 0.57), and AFP (P = 0.28). While, KL-6 (P < 0.001) and CYFRA21-1 (P = 0.006) had publication bias (Fig. 4). We further adjusted for publication bias using the Trim and Fill method and found that CYFRA21-1 elevated in RA- ILD patients (SMD = 0.475, 95%CI [0.058–0.892], p = 0.026) (Fig. 2). While, the results of KL-6 were consistent with previous findings, showing significant elevation in RA-ILD. Funnel plot for CYFRA21-1 (A and B) and KL-6 (C and D) Diagnostic performance of tumor markers in distinguishing RA-ILD from RA-non-ILD Based on the diagnostic accuracy metrics reported in the original studies, tumor markers demonstrated significant discriminatory ability for distinguishing RA-ILD from non-ILD RA patients (P < 0.001). Among the reported findings, KL-6 showed the highest AUC value (AUC = 0.964, 95% CI = 0.908–1.000) with an optimal cutoff of 376.8 U/mL (sensitivity = 93.3%), followed by CYFRA21-1 (AUC = 0.906, cutoff = 3.40 ng/mL, sensitivity = 75.0%). These values represent study-level findings as originally reported and are presented descriptively without pooling. Other markers showed moderate performance: CA199 (AUC = 0.796, cutoff = 13.42 U/mL, sensitivity = 87.0%), CA153 (AUC = 0.799, cutoff = 23.43 U/mL, sensitivity = 66.7%), CEA (AUC = 0.758, cutoff = 2.89 ng/mL, sensitivity = 63.6%), and CA125 (AUC = 0.617, cutoff = 19.93 U/mL, sensitivity = 88.9%) (Fig. 5). Discussion This systematic review and meta-analysis, encompassing 36 studies and 7,300 RA patients, provides compelling evidence for the diagnostic and pathogenic relevance of tumor markers in RA-ILD. Our findings establish a hierarchy of effect sizes: HE4 and CA153 demonstrated large effects (SMD > 0.8), KL-6 and CA125 showed moderate-to-large effects, while NSE exhibited a small but significant elevation. In terms of diagnostic performance, KL-6 (AUC = 0.964) and CYFRA21-1 (AUC = 0.906) emerged as the most robust discriminators of RA-ILD. These results carry significant implications for discriminating RA-ILD from RA-non-ILD strategies, pathophysiological insights, and clinical management of this life-threatening complication. KL-6, a glycoprotein expressed on alveolar type II pneumocytes and bronchiolar epithelial cells, has been extensively studied in ILD due to its role in epithelial repair and fibrosis. Its exceptional sensitivity, large effect size (SMD = 1.68), and highest reported AUC (0.964) in this analysis align with its established function as a marker of alveolar epithelial injury. However, it should be noted that this AUC reflects study-level findings and has not been pooled through a formal diagnostic meta-analysis. Elevated KL-6 levels in RA-ILD likely reflect subclinical pulmonary damage preceding radiographic changes, making it a critical tool for early intervention. Mechanistically, KL-6 may directly contribute to fibrogenesis by promoting fibroblast proliferation via the PI3K/Akt pathway, as demonstrated in idiopathic pulmonary fibrosis (IPF) models. This dual role as both a biomarker and a potential mediator of fibrosis underscores its clinical relevance [40]. CYFRA21-1, a soluble fragment of cytokeratin 19, similarly showed favorable discriminatory ability in the reported studies, with a modest effect size (SMD = 0.48) but a high reported AUC (0.906) [41]. Its elevation in RA-ILD patients, even after adjusting for publication bias, suggests its utility in detecting epithelial apoptosis—a hallmark of progressive lung injury. Unlike KL-6, which is specific to alveolar epithelium, CYFRA21-1’s broader expression in bronchial epithelium may explain its higher specificity in distinguishing RA-ILD from non-ILD RA patients. The complementary roles of KL-6 (sensitivity-driven) and CYFRA21-1 (specificity-driven) advocate for their combined use in multi-marker panels to optimize diagnostic accuracy. The moderate performance of CA199 (SMD = 0.90) and CA153 (SMD = 1.70) highlights their association with systemic inflammation and mucin hypersecretion, processes implicated in both RA and ILD. CA199, a sialylated Lewis antigen, is overexpressed in inflamed tissues and may reflect endothelial activation in pulmonary vasculature [42]. CA125, despite a large effect size (SMD = 1.10), showed a lower AUC (0.617) but demonstrated high sensitivity, suggesting utility as a “rule-out” test. Its elevation may reflect mesothelial injury or pleural involvement, common in advanced RA-ILD. However, the variability in CA125’s diagnostic performance across studies underscores the influence of confounding factors such as concurrent ovarian pathology or cardiovascular comorbidities, necessitating cautious interpretation in clinical practice [43]. Significant heterogeneity was observed in the meta-analyses of CA125, CA199, CEA, KL-6, AFP, and CA153. Our pre-specified subgroup analyses (Table 4) suggested that sample size and geographic region may contribute to this variability. For CA199 and CA153, heterogeneity was notably lower in studies with larger sample sizes (≥ 100), indicating that smaller studies may have inflated effect estimates. For KL-6, studies conducted in non-Chinese populations showed larger effect sizes but also higher heterogeneity, possibly reflecting differences in disease severity, assay methods, or ethnic variations in baseline KL-6 levels [44]. The persistent heterogeneity across subgroups suggests that other unmeasured factors may also play a role. For markers such as CA125 and CA199, where heterogeneity remained extremely high (I² > 90%), the choice of assay platform (e.g., ELISA vs. chemiluminescence) is a potential contributor, as different platforms have distinct analytical sensitivities, reference ranges, and antibody specificities. Additionally, variations in RA disease activity across cohorts may significantly influence serum concentrations of these biomarkers, given that systemic inflammation can upregulate tumor‑associated antigen expression. Unfortunately, the included studies did not uniformly report disease activity scores or assay platform details, precluding formal meta‑regression. Other potential sources include differences in patient selection criteria and unadjusted confounders. Standardizing assays, adjusting for disease activity in future primary studies, and establishing population‑adjusted reference ranges are imperative to enhance reproducibility and clinical applicability. The Trim and Fill adjustment validated CYFRA21-1’s association with RA-ILD, reinforcing its reliability despite initial publication bias. Conversely, KL-6’s results remained stable post-adjustment, attesting to the robustness of its association. However, the persistent bias in KL-6 studies may reflect selective reporting of positive findings in smaller cohorts, emphasizing the need for prospective validation in larger, diverse populations. While this analysis clarifies the diagnostic utility of major biomarkers, several limitations warrant attention. First, the predominance of retrospective studies introduces risks of selection and recall bias. Second, the lack of longitudinal data limits insights into biomarker dynamics during disease progression or treatment response. Third, several biomarkers such as YKL-40 and Pro GRP were reported in fewer than three studies, precluding definitive conclusions—a notable gap especially given YKL-40’s established role in lung inflammation. Fourth, the summary of diagnostic accuracy metrics (AUC, sensitivity, specificity) was based on data reported in the original studies rather than a formal diagnostic test accuracy meta-analysis, which limits our ability to derive pooled estimates of sensitivity and specificity. Furthermore, the geographic distribution of included studies showed a strong predominance of cohorts from China (32 out of 36), which may affect the generalizability of our findings to other ethnic and regional populations. Fifth, despite subgroup analyses, substantial heterogeneity remained for several biomarkers, indicating that unmeasured clinical or methodological factors continue to influence the pooled estimates. Future prospective, multicentric studies across diverse settings are needed to validate these biomarkers and establish standardized assay protocols. Finally, the diagnostic accuracy data were summarized descriptively without formal hierarchical meta-analysis; thus, the reported AUC values should be interpreted as exploratory study-level findings rather than pooled estimates. Prospective multicenter studies should prioritize three objectives. First, develop consensus guidelines for biomarker cutoff standardization, accounting for ethnic, age-related, and comorbidity-driven variability. Second, integrate biomarkers with multi-omics data to elucidate their roles in RA-ILD pathogenesis. For instance, KL-6’s interaction with fibroblast growth factor receptors merits investigation as a therapeutic target. Third, validate multi-marker panels against HRCT and pulmonary function tests in pre-symptomatic RA patients, enabling risk stratification and personalized monitoring. Conclusions This systematic review and meta-analysis of 36 studies comprising 7,300 RA patients demonstrates that several tumor-associated biomarkers are significantly elevated in RA-ILD. CA125, CA199, CA153, CEA, KL-6, NSE, HE4, and CYFRA21-1 were identified as significantly associated biomarkers between RA-ILD and RA-non-ILD. Among these, KL-6 and CYFRA21-1 exhibited the highest diagnostic accuracy, with large effect sizes and favorable AUC values, supporting their potential utility in discriminating RA-ILD from RA-non-ILD and risk stratification. These findings provide a serological basis for incorporating selected tumor markers into multidimensional screening strategies for RA-ILD. Importantly, while KL-6 and CYFRA21-1 show high diagnostic potential, they should currently be viewed as screening “red flags” to prompt HRCT imaging rather than replacements for radiological diagnosis. Prospective validation and assay standardization are warranted before routine clinical implementation. Supplementary Information Below is the link to the electronic supplementary material. Supplementary Material 1 Figure2-CA125 Figure2-CA153 Figure2-CA199 Figure2-CEA Figure2-CYFRA12-1 Figure2-KL-6 Figure2-HE4 Figure2-NSE Acknowledgements I would like to express gratitude to Liang Liu in the Second People’s Hospital of Qinzhou and Liu Yudong in Peking University People’s Hospital for providing the raw data for my meta-analysis. Author contributions Wei Qijiao designed this topic and drafted the article. Wang Na approved the final version of manuscript. Zeng Qiaoqian and Wu Qing screened the title and abstracts, extracted the data and assessed the quality of the included studies independently. Wei Qijiao decided any disagreement. Li Bowen helped to make the search strategy. Funding This work was supported by National Natural Science Foundation of China(82205189), a grant from Shanghai Pujiang Young RheumatologistsTraining Program (SPROG2403) , the Development of Fudan Children's Flagship Hospital for Integrated Chinese and Western Medicine (2021-2023)-0205-03 and 2025 General Hospital Chinese-Western Medicine Collaborative Guidance Project (ZHYYZXYYD-202503). Data availability The datasets generated during and/or analyzed during the current study are available from the corresponding author on reasonable request. Declarations Ethics approval and consent to participate Not applicable. Human ethics Not applicable. Consent for publication Not applicable. Competing interests The authors declare no competing interests. Footnotes Publisher’s note Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations. Zeng Qiaoqian, Wu Qing and Li Bowen contributed equally to this work. Contributor Information Wei Qijiao, Email: weiqijiao@126.com. Wang Na, Email: 18203638814@163.com. References Associated Data Supplementary Materials Supplementary Material 1 Figure2-CA125 Figure2-CA153 Figure2-CA199 Figure2-CEA Figure2-CYFRA12-1 Figure2-KL-6 Figure2-HE4 Figure2-NSE Data Availability Statement The datasets generated during and/or analyzed during the current study are available from the corresponding author on reasonable request.